Using Stream Channel Reference Data to Guide Land Management Decisions
Pete Bengeyfield
Beaverhead-Deerlodge National Forest, Montana
In order to make informed land management decisions, managers concerned with aquatic resources need
information at a variety of scales. “Project level” decisions require fine-scale information relating to specific
small watersheds or stream reaches, while planning level decisions require a characterization of broad scale
patterns that will influence allocation and scheduling of activities. Quantified reference data can provide a
basis for decisions at both scales by defining conditions that would exist in stream channels under the present
climate and tectonic regimes in the absence of land management. At smaller scales, data from undisturbed
reaches can be directly compared with similar managed reaches to determine departures from reference
conditions, while at larger scales reference data from a number of reaches can be use to define desired
conditions across a region.
Keywords: reference reach, TMDL, forest planning, sediment, channel morphology
PHYSICAL SETTING
Most of the data for this study were collected on the
Beaverhead-Deerlodge National Forest (B-D), covering
approximately 3.4 million acres (1.4 million hectares)
in southwest Montana. The region is bordered by the
continental divide on the west, and contains the headwaters
of four major river systems (Big Hole, Beaverhead, Madison
and Ruby). Wide valley bottoms are separated by isolated
mountain ranges. The valley bottoms of the major rivers
are mostly in private ownership, with public land located
at higher elevations.
The dominant physical realities of this region are
high elevation and low precipitation. Valley bottoms
are semi-arid, averaging about 5800 feet (1769 m) in
elevation and receiving an average of 13.6 inches (34.5
cm) of precipitation a year. The mountains receive more
precipitation (27.2 inches [69.1 cm] per year), with most
of that in the form of snow. Overall, annual precipitation
has averaged 20.4 inches (51.8 cm) since 1980.
Livestock grazing is the land use that most affects
aquatic values on the Beaverhead-Deerlodge National
Forest. Seventy-eight percent of the forest is located in
grazing allotments. Roadless and wilderness areas account
for 62% of the forest, so some of the more common
effects to aquatics from roads, timber harvest, and mining
are either isolated or minimal. There is a wide range of
suitable rangeland contained within the allotments. Some
allotments, mainly on the eastern portion of the forest, are
M Furniss, C Clifton, and K Ronnenberg, eds., 2007. Advancing
the Fundamental Sciences: Proceedings of the Forest Service National
Earth Sciences Conference, San Diego, CA, 18-22 October 2004, PNWGTR-689, Portland, OR: U.S. Department of Agriculture, Forest
Service, Pacific Northwest Research Station.
predominately suitable, while others are heavily timbered
with little suitable range.
Additional data were collected throughout the Greater
Yellowstone Ecosystem, which stretches south and west
of the Beaverhead-Deerlodge into Wyoming and Idaho.
Some of the Ecosystem mimics conditions on the B-D,
although precipitation levels are often somewhat higher.
METHODS
A comprehensive stream channel survey was initiated on
the Beaverhead-Deerlodge National Forest in 1991. The
survey was designed to collect stream channel data that
would characterize physical channel attributes at the reach
scale. Survey sites were chosen to: 1) identify reference
conditions; 2) identify problem areas for restoration; and
3) portray the range of stream conditions throughout the
forest. The following parameters were measured as part of
the standard survey:
1. A monumented cross-section depicting floodplain,
bankfull elevation, water surface, and thalweg. The cross
section is used to compute entrenchment and width-todepth ratio. (Rod and level)
2. Sinuosity (Pacing)
3. Gradient (Rod and level)
4. Stream substrate (Wolman pebble count, Wolman
1954)
5. Fifty bankfull widths and depths (measuring rod)
6. Channel Stability Evaluation (Pfankuch 1975)
7. Bank Erosion Hazard Index (Rosgen 1996)
8. Valley Bottom Width (pacing)
9. Reach photographs
10. Notes describing overall conditions with respect to
previous and current management effects.
70
STREAM CHANNEL REFERENCE DATA AND MANAGEMENT
The sites are permanently established with painted rebar
and surveyed to local benchmarks, usually a nail in a
tree. Where applicable, measurement protocols given in
Harrelson et al. (1994) were used.
By 2002, a total of 682 stream reaches had been
measured throughout the forest. Forty-four percent of the
total streams on the forest have at least one survey, with
many streams having multiple sites. On average, 3.3 survey
sites were measured per 6th-code HUC (Hydrologic Unit,
or watershed). Measured sites were considered to be a
“representative reach” (Gordon et al. 1992), and depict
conditions along a greater length of stream channel. The
survey sites depict conditions on 3.7% of the stream length
on the forest. Of the 682 sites, 137 were determined to
be in reference condition. These have been combined with
an additional 78 reference sites measured throughout the
Greater Yellowstone Area (GYA) in the summer of 2002
to create a reference dataset of 215 reaches. The dataset
includes stream reaches that represent watersheds of a
variety of sizes, geologies, and precipitation ranges.
THE USE OF REFERENCE REACH DATA IN PROJECT
PLANNING (NEPA ANALYSIS)
Under the National Environmental Policy Act (NEPA),
federal agencies initiating land management activities must
complete an analysis to determine the effect of the project
on the natural environment. The Affected Environment
section of the NEPA analysis requires a determination of
the existing condition before the effect of the proposed
project is evaluated. One way to evaluate existing condition
is to compare current conditions with reference data for
similar sites (Frissell et al. 1986). The stream surveys
enabled us to do this for the Beaverhead-Deerlodge
National Forest.
The concept of using minimally disturbed sites as
references has appeared in the literature in recent years
(Dissmeyer 1994; Minshall 1994; Maxwell et al. 1995).
Recent methodologies for analyzing watershed conditions
(USDA et al. 1994, 1995; McCammon et al. 1998)
recommend the use of reference watersheds as a means of
determining the effects of land management
Reference reaches are matched with project reaches
on the basis of the similarity of their valley bottom
widths, valley bottom gradient, and the drainage area
above the reach. Rosgen (1996) demonstrated that valley
types provide an indication of stream channel morphology.
Bengeyfield (1999) showed the relationship between valley
features and drainage area to stream types for southwest
Montana. Using a Classification and Regression Tree
procedure (Brieman et al. 1984) and an earlier version of
this data set, he used valley bottom width, valley bottom
gradient, and drainage area to predict Level One stream
types. A cross-validation procedure showed that E stream
types, for example, were correctly predicted 89% of the
time. Analyzing the data from the reference data set and
project reach stream surveys produces a comparison table
(Table 1).
Table 1. Comparison of stream survey parameters between
project reach (Lone Tree Creek) and reference reach (Sunshine
Creek). * indicates data not collected.
Lone Tree
Creek
Sunshine
Creek
25
75
Valley Bottom Gradient (VBG) (%)
2.53
2.75
Area (acres)
1500
704
Stream Type
B4
E4
Entrenchment
1.96
9.63
Width/Depth Ratio
28.5
3.3
Sinuosity
1.05
1.8
Gradient (%)
2.41
1.53
30
13
Parameter
Valley Bottom Width (VBW) (ft)
D50 (mm)
W50 (ft)
Channel Stability
BEHI
3.7
5.9
64 - Fair
48 - Good
*
24.8
A comparison of the data for Lone Tree Creek (project)
and Sunshine Creek (reference) shows that both valley
bottoms are narrow and moderately steep, and their
watersheds are small (Table 1). However, they contain
different stream types (B4 vs. E4). Lone Tree Creek is more
entrenched (1.96 vs. 9.63), wider and shallower (widthto-depth ratio = 28.5 vs. 3.3), straighter (sinuosity = 1.1
vs. 1.8), and steeper (gradient = 2.4 vs. 1.5) than the
reference. It is not only wider at the cross section, but
along the entire reach (W50 = 5.9 vs. 3.7). Lone Tree Creek
is not as stable as Sunshine Creek. (Channel Stability =
64 vs. 48), and is more susceptible to streambank erosion
(Bank Erosion Hazard Index [BEHI] = 30.5 vs. 24.8). The
relationships between the project and reference reaches can
be displayed graphically as shown in Figures 1 and 2.
An analysis of the data and graphs shows Lone Tree
Creek differing from the reference for most of the stream
parameters measured. The largest differences are in channel
dimension, with somewhat smaller differences in pattern
and profile. Lone Tree Creek is wider, straighter, and
steeper than Sunshine Creek. Channel Stability and Bank
Erosion Hazard reflect similar patterns. Notes taken onsite
at the time of the survey indicate heavy trampling by
livestock, willows that are in poor condition, and a species
of sedge that increases with grazing pressure.
BENGEYFIELD
Figure 1. Cross-sections of Sunshine Creek (reference) and Lone
Tree Creek (project).
71
allowed departure from desired conditions to be displayed.
By including these data in Chapter 3 of the NEPA
document, the need for change was established, and an
alternative was generated that prescribed more stringent
standards for livestock management.
THE USE OF REFERENCE REACH DATA IN A REGULATORY
FRAMEWORK (TMDL ANALYSIS)
By comparing onsite data from Lone Tree Creek to
reference conditions, the departure from expected channel
dimensions can be quantified and displayed graphically. In
this case, the notes describe a disturbance (livestock grazing)
that would logically produce the channel conditions
measured, suggesting a cause/effect relationship for the
existing condition. Similar effects due to grazing have
been documented by Clifton (1989), Clary and Webster
(1990), and Kovalchik and Elmore (1992). The data and
analysis suggest there is considerable difference between
the existing and desired conditions in Lone Tree Creek,
necessitating a change in grazing management in order to
restore reference conditions.
Lone Tree Creek is one of five reaches surveyed for the
Antelope Creek Grazing Allotment. Employing reference
data, site specific conditions were described for individual
reaches and then compared to the reference data. This
Figure 2. Cumulative distribution
curve for bankfull stream widths for
Sunshine Creek (reference) and Lone
Tree Creek (project).
In recent years, states have identified Water Quality
Limited Segments (WQLS) as impaired water bodies that
are in need of restoration in order to meet the requirements
of the Clean Water Act. This process is designed to be
quantitative, and specific levels of a pollutant are required
to be met before the stream can be removed from the
impaired list. The use of reference conditions is considered
to be an acceptable concept for determining target levels of
pollutants (EPA 1999). Furthermore, it is recommended
that reference conditions be measured across a region to
define variability (Frissel et al. 1986; Minshall 1994).
Data from the measurement of 215 reference
reaches throughout southwest Montana and the Greater
Yellowstone Ecosystem provides information that can be
used to establish acceptable target levels for specific Rosgen
stream types (Rosgen 1996). Figures 3 through 5 display
cumulative distribution curves for substrate size and width/
depth ratio derived from the reference data set for stream
types B, C, and E.
For example, if a stream were placed on a state’s 303(d)
list as being impaired for sediment, and the impaired reach
were a B4 stream type, then Figure 3 could be used to
establish target levels for substrate distribution. If fine
sediment were a concern, the target might be that only 16%
of the substrate could be below 6 mm in size. If bedload
transport were a concern, the D84 of the impaired reach
might be targeted at 160 mm. Confidence intervals define
acceptable range of variability for monitoring purposes.
If an E stream were placed on the 303(d) list for
habitat alteration as a result of streambank alteration due
72
STREAM CHANNEL REFERENCE DATA AND MANAGEMENT
Figure 3. Particle size distribution, B stream
types, with 95% confidence limits.
Figure 4. Particle size distribution, C stream
types, with 95% confidence limits.
Figure 5. Particle size distribution, E stream
types, with 95% confidence limits.
Table 2. Distribution of stream types in Beaverhead/Greater
Yellowstone Area dataset.
Level 2 Stream Type
B Streams C Streams E Streams
2
1
3
26
12
10
4
16
22
65
2
18
5
Ea
25
BENGEYFIELD
to livestock grazing, then a target might be established for
width/depth ratio in order to restore overhanging banks.
The distribution throughout the reach should be such that
the average width/depth ratio should be about 3 to 3.5.
THE USE OF REFERENCE REACH DATA AT THE LANDSCAPE
SCALE (FOREST PLANNING)
National Forests throughout the country are presently
going through a revision of Forest Plans that were first
completed in the 1980s. As part of that process, the
Desired Future Condition (DFC) for individual resources
must be described (CFR 219). For stream channels, a
logical DFC would be to maintain the dimension, pattern,
and profile that have been developed by natural flow
regimes under the present climate and tectonic regimes.
Reference reach data can help in describing these conditions
73
and establishing goals for management.
The reference reach dataset for the BeaverheadDeerlodge and Greater Yellowstone Ecosystem contains
215 reaches. These were stratified by stream type, with
results summarized in Table 2. The B2 and C5 strata were
dropped from further analysis because their sample size
was too small. This left a reference dataset of 213 reaches.
The remaining strata were subject to statistical analysis for
mean, standard deviation, and 90% confidence interval
(Table 3).
These data localize the Rosgen classification and can
form the basis for describing DFC for each of the stream
types. For example, for E stream types to maintain stable
channels given the natural flow regime, they should have
a width/depth ratio close to the mean for E stream types,
in this case, 6. If the width/depth ratio of a given reach
fell between 5.5 and 6.5, a manager could be 90%
Table 3. Statistics for reference reach strata representing various stream types. Confidence Intervals are all α = 0.1. W/D ratio is
width to depth ratio; % < 6mm is the percent of the substrate smaller than 6 mm in diameter; D50 is median particle size; D84 is the
particle size 84% of the substrate is smaller than; W50 is average channel width; CS is the channel stability rating (Pfankuch 1975)
; BEHI is the Bank Erosion Hazard Index (Rosgen 1996).
Stream
Type Statistic Entrenchment W/D ratio
Mean
SD
90% CI
B4
Mean
N=16
SD
90% CI
Total B Mean
N=42
SD
90% CI
C3
Mean
N=12
SD
90% CI
C4
Mean
N=22
SD
90% CI
Total C Mean
N=34
SD
90% CI
E3
Mean
N=10
SD
90% CI
E4
Mean
N=65
SD
90% CI
E5
Mean
N=18 SD
90% CI
Ea
Mean
N=25 SD
90% CI
Total E Mean
N=118 SD
90% CI
Sinuosity
Gradient % < 6mm
B3
N=26
6
3-8
17
1.7
0.3
1.78 - 1.62
13.3
1.22
0.039
5.9
0.14
0.026
14.8 - 11.8 1.26 - 1.18 0.049-0.029
12 - 22
5
0 - 10
16
18.2
16.6
22.8 - 13.6
23
1.48
0.013
11.9
0.35
0.014
26.2 - 19.8 1.57 - 1.39 0.017-0.009
13 - 18
8
0 - 18
22
19 - 25
64
11.1
14.4
13.3 - 8.9
6
3.4
6.5 - 5.5
1.25
0.062
0.21
0.022
1.32 - 1.18 0.069-0.055
1.6
0.017
0.4
0.01
1.66 - 1.54 0.019-0.015
60 - 71
D50
D84
107
41
120-94
41
18
48-34
306
195
369-243
148
70
177-119
91
29
104-78
39
14
44 - 34
206
84
244-168
104
53
133 - 85
82
22
93 - 71
25
16
28 - 22
201
79
160
70
45
79 - 61
38.5
73.8
67.1- 9.9
0.48
0.54
0.69-0.27
W50
CS
BEHI
13.4
64.7
25.5
6.5
16.1
7.4
15.5-11.3 69.6-59.8 27.9-24.1
29.4
71.1
24.3
19.5
18.8
6.8
41.4-17.4 82.5-59.7 28.5-20.1
6.6
6.1
8.6-4.6
62.5
20.9
15.3
4.9
67.6-57.4 22.7-19.1
74
STREAM CHANNEL REFERENCE DATA AND MANAGEMENT
certain it was representing that mean. By describing each
of the parameters in these terms, the DFC for that
particular stream type can be described. How these
data are eventually displayed in the Forest Plan will be
determined in an interdisciplinary mode. For example,
in the Rosgen classification, reaches that are slightly
entrenched really have no upper limit on entrenchment
ratio. Therefore, establishing a range of 13.6 to 22.8
for C reach entrenchment would be unrealistic and
unnecessarily constraining. In this case, the DFC for C
reach entrenchment might be better expressed as > 13.6.
DISCUSSION
The measurement of a variety of physical parameters in
watersheds that are undisturbed or minimally disturbed
can be used to define physical stream conditions that reflect
the present climate and tectonic regime of a region. If these
measurements are used as references, land managers can
estimate the dimension, pattern, and profile that would be
the result of natural processes for stream channels within
their area. By measuring the parameters that are used
as delineative criteria in the Rosgen classification, stream
types can be locally defined. For example, E stream types
are defined in the classification system as having a width/
depth ratio of less than 12. In the Greater Yellowstone
Ecosystem/southwest Montana area, measurements of 118
E stream reaches show an average width/depth ratio of 6.
Consequently, if a given E reach had a width/depth ratio
of 10, and the stream banks were trampled by livestock,
that reach might be determined to be trending away from
stability as a result of getting wider and shallower.
These data can be used at various scales depending
on the situation. For NEPA documents, where very site
specific data is required to characterize stream conditions,
reference and project reaches may be matched based on
valley bottom characteristics. Demonstrating the difference
between reference and project conditions will give the
manager an idea of the status and trend of streams in the
project area and how they may respond to a proposed
action. On the Beaverhead-Deerlodge National Forest,
NEPA documents have incorporated this approach since
1997.
In the TMDL arena, targets for parameters that define
the condition of a stream with respect to beneficial uses
must be established. Reference data, sampled throughout
the area, can help define the conditions whereby beneficial
uses can be expected to be self-maintaining. In Montana,
the Ruby River and Yaak River TMDLs are presently using
reference data to help establish targets for sediment and
channel dimensions. At the Forest Planning scale, Desired
Future Conditions define the hoped for end result of a suite
of management operations over the entire forest. In past
planning efforts, the DFCs for streams were often simple
narratives that were neither quantitative nor measurable.
The ability to gather reference data over a wide geographic
area and a variety of stream types allows the DFC for each
stream type to be described in a manner that provides the
land manager a picture of the conditions that should exist
regardless of the suite of management activities applied
to the land. In the Forest Plan Revision process, the
Beaverhead-Deerlodge National Forest is basing its DFCs
for aquatics on local reference data. Across all scales, the
ability to quantify standards, targets, and DFCs, makes
any subsequent monitoring program more effective.
Reference data can provide land managers timely and
excellent information with which to make decisions.
However, they are not a panacea, and should not be applied
in a vacuum. The need for professional interpretation at
all levels of analysis is paramount. Choosing individual
reference reaches, setting confidence limits, establishing
acceptable risk, and determining the relative importance
of different types of reference data, all require field-based
knowledge of the watersheds in question. Rather than
dogmatically establishing limits to management, reference
data should be looked on as a tool that enables specialists
to make better recommendations, and allows managers to
gain a better understanding of how aquatic systems work.
LITERATURE CITED
Bengeyfield, P. 1999. Analysis of reference reach data and the
prediction of stream types. In: D S Olsen and JP Potyondy
(eds.), Wildland hydrology: proceedings, specialty conference,
June 30-July 2, 1999, Bozeman, MT. Technical publication
series, TPS-99-3. Herndon, VA: American Water Resources
Association: 245-254.
Breiman, L, J Friedman, R Oshen, and C Stone. 1984.
Classification and regression trees. Pacific Grove:
Wadsworth.
Clary, WP, and BF Webster. 1990. Riparian grazing guidelines
for the Intermountain Region. Rangelands 12(4): 11.
Clifton, C. 1989. Effects of vegetation and land use on channel
morphology. Billings, MT BLM-MT-89-001-4351.
Code of Federal Regulations [CFR] 219. 1999. Vol. 36, Parts
200-299. Washington, DC: National Archives and Records
Administration, Office of Federal Register: 53-79.
Dissmeyer, GE. 1993. Evaluating the effectiveness of forestry
best management practices in meeting water quality goals
or standards. Misc. Pub. 1520. Atlanta, GA: USDA Forest
Service, Southern Region, 166 p.
Frissell, CA, WJ Liss, CE Warren, and MD Hurley. 1986.
A hierarchical framework for stream habitat classification:
BENGEYFIELD
Viewing streams in a watershed context. Environmental
Management 10(2): 199-214.
Gordon, ND, TA McMahon, and BL Finlayson. 1992. Stream
hydrology: An introduction for ecologists. New York: John
Wiley & Sons.
Harrelson, CC, CL Rawlins, JP Potyondy. 1994. Stream channel
reference sites: an illustrated guide to field technique. Gen.
Tech. Rep. RM-GTR-245. Fort Collins, CO: US Forest
Service, Rocky Mountain Forest and Range Experiment
Station. 61 p.
Kovalchik, BL and W Elmore. 1992. Effects of cattle grazing
systems on willow dominated plant associations in central
Oregon. In: WP Clarry, ED McArthur, D Beduna, CL
Wanboldt (eds) Proceedings - Symposium on ecology and
management of riparian shrub communities. Gen. Tech.
Rep. INT-GTR-289, Sun Valley, ID: USDA Forest Service,
Intermountain Research Station: 112-119.
Maxwell, JR, CJ Edwards, ME Jensen, S Paustian, H Parrott,
and DM Hill. 1995. A hierarchical framework of aquatic
ecological units in North America (Nearctic Zone). Gen.
Tech. Rep. NC-GTR-176. St. Paul, MN: USDA Forest
Service, North Central Research Station.
McCammon, B, J Rector, and K Gebhardt. 1998. A framework
for analyzing the hydrologic condition of watersheds. BLM
Technical Note 405. Denver, CO: USDA Forest Service and
US Dept. of the Interior, Bureau of Land Management,
National Applied Resources Science Center, 37 p.
75
Minshall, GW. 1994. Stream-riparian ecosystems: Rationale and
methods for basin level assessments of management effects.
Gen. Tech. Rep. PNW-GTR-318. Portland, OR: USDA
Forest Service, Pacific Northwest Research Station.
Pfankuch, DJ. 1975. Stream reach inventory and channel stability
evaluation. USDA Forest Service, Rl-75-002, #696-260/200.
Washington, DC: Government Printing Office, 26 p.
Rosgen, DL. 1996. Applied river morphology. Lakewood, CO:
Wildland Hydrology.
U.S. Environmental Protection Agency. 1999. Protocol for
developing sediment TMDLs. EPA 841-B-99-004, Office
of Water (4503F) Washington, DC: U.S. Environmental
Protection Agency.
U.S. Department of Agriculture et al. 1994. A federal agency
guide for pilot watershed analysis. Version 1.2. Portland, OR:
USDA Forest Service.
U.S. Department of Agriculture et al. 1995. Ecosystem analysis
at the watershed scale: federal guide for watershed analysis.
Version 2.2. Portland, OR: USDA Forest Service, Regional
Ecosystem Office, 26 p.
Wolman, MG. 1954. A method of sampling coarse river bed
material. Transactions of the American Geophysical Union 35(6):
951-956.